tnetstring shootout

README.rst

I was a bit skeptical about a new data format call tnetstring. It claims to be both human and machine readable. They claim that a parser is easier to built for a tnetstring and therefore is better. In fact, while the implementation of the parser may be easier, human readability suffers with negligible speed gains.

I assert that if human readability is harmed in favor of speed, then you would be much better off using a binary solution such as Protobufs.

Results

Time is in milliseconds:

(experiments)emoritzmbpro:tnetstringshootout emoritz$ python tnet.py
13.1070613861
(experiments)emoritzmbpro:tnetstringshootout emoritz$ python jsontest.py
1780.90310097
(experiments)emoritzmbpro:tnetstringshootout emoritz$ python json-min-test.py
1514.28318024
(experiments)emoritzmbpro:tnetstringshootout emoritz$ python simplejsontest.py
35.9201431274
(experiments)emoritzmbpro:tnetstringshootout emoritz$ python simplejson-min-test.py
16.233921051
(experiments)emoritzmbpro:tnetstringshootout emoritz$ python pkltest.py
36.780834198
(experiments)emoritzmbpro:tnetstringshootout emoritz$ python bsontest.py
22032.9511166

bsontest.py

from data import data
import bson

from time import time
doc = bson.dumps(data)

start = time()
result = bson.loads(doc)
end = time()
print (end - start)*1000

data.py

# Create a large nested dictionary


root = {}
current = root
for i in range(100):
    current["big list"] = range(1000)
    current["big string"] = "a" * 1000
    current['child'] = {}
    current = current['child']

data = root

json-min-test.py

from data import data
import json

from time import time
doc = json.dumps(data)

start = time()
result = json.loads(doc)
end = time()
print (end - start)*1000

jsontest.py

from data import data
import json

from time import time
doc = json.dumps(data, indent=4)

start = time()
result = json.loads(doc)
end = time()
print (end - start)*1000

pkltest.py

from data import data
import cPickle as pickle

from time import time
doc = pickle.dumps(data)

start = time()
result = pickle.loads(doc)
end = time()
print (end - start)*1000

simplejson-min-test.py

from data import data
import simplejson as json

from time import time
doc = json.dumps(data)

start = time()
result = json.loads(doc)
end = time()
print (end - start)*1000

simplejsontest.py

from data import data
import simplejson as json

from time import time
doc = json.dumps(data, indent=4)

start = time()
result = json.loads(doc)
end = time()
print (end - start)*1000

tnet.py

from data import data
import tnetstring
from time import time
doc = tnetstring.dumps(data)

start = time()
result = tnetstring.loads(doc)
end = time()
print (end - start)*1000